To date, the precise role of DC/L-SIGN in SARS-CoV infections is unclear and controversial. In a number of studies, DC/L-SIGN have been reported to function as infection enhancer factors but not as receptors (30
); while these lectins capture and transfer viruses to neighboring target cells expressing ACE2, thereby facilitating infections in trans
, they themselves did not appear to support virus infections directly. In contrast to these reports, Jeffers et al. (18
) have shown that L-SIGN could serve as an alternative receptor. In agreement with the latter study, the results of our study clearly demonstrated that both DC-SIGN and L-SIGN do indeed serve as receptors independently of ACE2, albeit less efficiently (Fig. ). The reason for the discrepancy is presently unknown.
One possible reason for the contradictory results is that different studies used different cell lines. While Jeffers et al. used CHO cells and we used HeLa cells, other studies used either quail-derived QT6 cells or B-THP-1 Raji B cells. If this speculation is true, then it raises the possibility that SARS-CoV uses a coreceptor, which could be present on CHO or HeLa cells but not on QT6 or B-THP-1 cells. Alternatively, DC/L-SIGN could be posttranslationally modified differently in different cell types, allowing the proteins on some cells to mediate virus infection but not those on others. Another possibility is that there could be subtle differences in membrane fluidity, which could affect the movement of membrane proteins within the lipid bilayer or membrane protein trafficking among different cell types. Regardless, additional studies will be necessary to better define the functional properties of DC/L-SIGN in SARS-CoV infections.
In this study, we have identified seven potential N-linked glycosylation sites that play an important role in L-SIGN-mediated SARS-CoV entry. They include residues N109, N118, N119, N158, and N227 within cluster I and N589 and N699 within cluster II. It should be mentioned that, to date, glycosylation has been positively confirmed physicochemically only for residues N118, N119, and N227 (6
) (Fig. ). However, asparagine residues at N109, N158, N589, and N699 are most likely glycosylated on the basis of the facts that (i) mutating these sites specifically diminished L-SIGN-mediated infections but not those mediated by ACE2, and (ii) mutating 10 other asparagine residues did not nonspecifically affect virus infectivity. Regardless of whether these asparagine residues are actually glycosylated or not, the results of our study indicate that these amino acids are critical for L-SIGN-mediated virus infections.
It is interesting that two glycosylation sites in cluster II are located distant from the five sites in cluster I (Fig. ). This result raises a question as to whether the glycosylation sites in the two clusters are indeed physically separated or whether they actually lie close together on a tertiary structure of the protein. Solving a crystal structure of an intact protein will be needed to address this question. It also is noteworthy that all seven glycosylation sites are situated clearly outside of the ACE2-binding domain (aa 318 to 510). In this regard, it was somewhat surprising that four glycosylation sites located close to or within the ACE2-binding domain (N269, N318, N330, and N357) were dispensable for DC/L-SIGN-mediated infections. Although we can speculate (see below), we do not yet know why only certain glycosylation sites are capable of mediating infections through DC/L-SIGN.
FIG. 9. Role of N-linked glycosylation sites in S-protein function. (A) A schematic diagram of S glycoprotein. Locations of seven glycosylation sites important for DC/L-SIGN-mediated infections are shown. Yellow lines indicate positions of 17 amino acids that (more ...)
In a recently reported study, it has been suggested that a region between amino acid residues 324 and 386 of S protein was the minimal DC-SIGN-binding domain and that glycosylation sites N330 and N357 are involved in DC-SIGN-mediated SARS-CoV infections (39
). This conclusion was based on the following observations. First, a recombinant baculovirus that expressed amino acid residues 17 to 386 of S protein fused to a truncated baculovirus envelope glycoprotein, gp64, was able to bind B-THP-1 cells expressing DN-SIGN but not a virus that expressed residues 17 to 324. Second, an epitope recognized by a monoclonal antibody that prevents binding of SARS-CoV to DC-SIGN was mapped to amino acid residues 363 to 368. Third, mutating N330 and/or N357 to glutamine partially reduced the infectivity of SARS-CoV pseudoviruses.
Despite seemingly contradictory results, it should be emphasized that the two studies are fundamentally different. In our study, we examined glycosylation sites that are critical for direct virus entry (i.e., using DC/L-SIGN only, independently of ACE2). In contrast, Shih et al. (39
) examined glycosylation sites important for capturing viruses by DC-SIGN-expressing B-THP-1 cells and subsequent trans
-infection of HepG2 target cells that express ACE2.
Besides the difference in the nature of the assay, we feel that Shih et al. (39
) may have overlooked the importance of other glycosylation sites for the following reasons. First, expression of amino acid residues 17 to 324 fused to baculovirus gp64 could have resulted in a misfolded protein. Therefore, to conclude simply that glycosylation sites between amino acids 17 and 324 are not involved in binding DC-SIGN is somewhat flawed. Second, the glycosylation pattern of S glycoprotein expressed in insect cells could be significantly different from that expressed in mammalian cells, in terms of both site usage and types of modifications. In fact, our recent mass spectrometry analyses of S1 protein expressed in insect cells using a recombinant baculovirus (4
) have shown that residues N227 and N269 are not glycosylated (unpublished data). Third, they utilized only a limited number of MAbs to identify ones that can block DC-SIGN-mediated trans
-infection. Finally, Shih et al. (39
) did not examine the effects of mutating other glycosylation sites.
According to our data, glycosylation sites at N330 and N357 do not support direct infections mediated by L-SIGN (Fig. ). However, these residues have been implicated in binding DC-SIGN and in mediating trans
). Together, these results seem to suggest that binding of DC/L-SIGN to glycans at these sites, unlike the seven sites we found to be important, does not lead to an eventual fusion of viral and cellular membranes. This failure could be due to certain geometrical constraints that prevent one of the post-receptor-binding events from occurring (e.g., conformational changes in S protein or insertion of a hydrophobic fusion domain into the cellular membrane). Examination of a crystal structure of the (ACE2) RBD shows that N330 and N357 are located distant from the RBM (Fig. ). N318, which also does not support direct infections mediated by DC/L-SIGN, is situated even further away (inferred based on the position of I319 on the crystal structure). These residues are located on a plane that is perpendicular to that of the RBM. One could speculate that such orientation prevents the protein from undergoing proper conformational changes. Determination of the structure of an intact protein (or at least the entire S1 domain) could facilitate better understanding of post-receptor-binding events of SARS-CoV entry processes.
Although the exact origin of SARS-CoV responsible for the 2002 and 2003 epidemic is unknown, molecular phylogenetic analyses and epidemiological studies indicate clear zoonotic transmission of the virus (11
). One likely source is Himalayan palm civets (Paguma larvata
). Amino acid sequence analyses of S glycoproteins of viruses isolated from civets (SZ3) and humans (Urbani) revealed a difference of 17 residues (Fig. ). Two of these changes, which lie within the RBM (K479N and S487T), have been shown to be critical for efficient binding to human ACE2 (25
). Plotting the mutations on a linear diagram of S glycoprotein revealed two major clusters of mutations (227 to 261 and 607 to 778), which suggests the importance of some of these changes for adaptations to grow in human cells (Fig. ). Interestingly, we noticed that mutations K227N and L701S have introduced two novel N-linked glycosylation sites at N227 and N699, both of which we found were important for L-SIGN-mediated infections (Fig. and ).
Extensive database searches of SARS-CoV S glycoproteins showed that K227N and L701S mutations are found in all viruses isolated from humans. Among viruses isolated from civets, there were strains with the L701S mutation alone (e.g., HC/SZ/61/03) or both K227N and L701S mutations (e.g., Civet014 or PC4-137). These viruses contained neither K479N nor S487T mutation. Some viruses contained both K227N and L701S mutations as well as K479N (e.g., PC4-115 isolated from a civet and GD03T0013 isolated from a human patient). From these observations, together with the fact that SZ3 is extremely poor in infecting cells using human ACE2, it is tempting to speculate that the two glycosylation sites at N227 and N699 have facilitated zoonotic transmission of SARS-CoV. As diagrammed in Fig. , SZ3 virus infection in humans might not be productive due to inefficient utilization of human ACE2 by the virus. However, a virus with glycosylation sites at N227 and/or N699 could replicate in humans better than SZ3 using DC/L-SIGN. This could provide opportunities for producing variants with K479N and/or S487T mutations, allowing the mutant viruses to efficiently utilize human ACE2. More detailed phylogenetic analyses using a greater number of virus isolates and additional site-directed mutagenesis studies will be required to provide conclusive answers.